Intradomain Topology and Routing

(Nick Feamster)

January 30, 2008

2

GeorgiaTech

Internet Routing Overview

• Today: Intradomain (i.e., “intra-AS”) routing• Monday: Interdomain routing

Comcast

Abilene

AT&T Cogent

Autonomous Systems (ASes)

3

Today: Routing Inside an AS

• Intra-AS topology– Nodes and edges– Example: Abilene

• Intradomain routing protocols– Distance Vector

• Split-horizon/Poison-reverse• Example: RIP

– Link State• Examples: OSPF, IS-IS

4

Key Questions

• Where to place “nodes”?– Typically in dense population centers

• Close to other providers (easier interconnection)• Close to customers (cheaper backhaul)

– Note: A “node” may in fact be a group of routers, located in a single city. Called a “Point-of-Presence” (PoP)

• Where to place “edges”?– Often constrained by location of fiber

5

Point-of-Presence (PoP)

• A “cluster” of routers in a single physical location

• Inter-PoP links– Long distances– High bandwidth

• Intra-PoP links– Cables between racks or

floors

PoP

6

Example: Abilene Network Topology

• Problem Set 1 will have a problem dealing with Abilene router configurations/topology.

7

Where’s Duke?

OC-48 (2.5GbpS uplink)NCREN Raleigh

8

Recent Development: NLR Packet Net

9

Problem: Routing

• Routing: the process by which nodes discover where to forward traffic so that it reaches a certain node

• Within an AS: there are two “styles”– Distance vector– Link State

10

Distance-Vector Routing

• Routers send routing table copies to neighbors• Routers compute costs to destination based on shortest

available path• Based on Bellman-Ford Algorithm

– dx(y) = minv{ c(x,v) + dv(y) }– Solution to this equation is x’s forwarding table

x y z

x 0 1 5

y

z

x y z

x

y 1 0 2

z

x y z

x

y

z 5 2 0

y

x z

1 2

5

11

Good News Travels Quickly

• When costs decrease, network converges quickly

x y z

x 0 1 3

y 1 0 2

z 3 2 0

x y z

x 0 1 3

y 1 0 2

z 3 2 0

x y z

x 0 1 3

y 1 0 2

z 3 2 0

y

x z

1 2

5

12

Problem: Bad News Travels Slowly

y

x z

1 2

50

60

x y z

x 0 60 50

y 5 0 2

z 3 2 0

x y z

x 0 60 50

y 5 0 2

z 7 2 0

Note also that there is a forwarding loop between y and z.

13

It Gets Worse

• Question: How long does this continue?• Answer: Until z’s path cost to x via y is greater than 50.

y

x z

1 2

50

60

x y z

x 0 60 50

y 5 0 2

z 3 2 0

x y z

x 0 60 50

y 5 0 2

z 7 2 0

14

“Solution”: Poison Reverse

• If z routes through y to get to x, z advertises infinite cost for x to y

• Does poison reverse always work?

x y z

x 0 1 3

y 1 0 2

z 3 2 0

x y z

x 0 1 X

y 1 0 2

z X 2 0

x y z

x 0 1 3

y 1 0 2

z 3 2 0

y

x z

1 2

5

15

Does Poison Reverse Always Work?

y

x z

1 3

50

60

w

1

1

16

Example: Routing Information Protocol

• Earliest IP routing protocol (1982 BSD)– Version 1: RFC 1058 – Version 2: RFC 2453

• Features– Edges have unit cost– “Infinity” = 16

• Sending Updates– Router listens for updates on UDP port 520– Message can contain up to 25 table entries

17

RIP Updates

• Initial– When router first starts, asks for copy of table for every neighbor– Uses it to iteratively generate own table

• Periodic– Table refresh every 30 seconds

• Triggered– When ever an entry changes, send copy of entry to neighbors

• Except for one causing update (split horizon rule)

– Neighbors use to update their tables

18

RIP: Staleness and Oscillation Control

• Small value for Infinity– Count to infinity doesn’t take very long

• Route Timer– Every route has timeout limit of 180 seconds

• Reached when haven’t received update from next hop for 6 periods

– If not updated, set to infinity– Soft-state

• Behavior– When router or link fails, can take minutes to stabilize

19

Link-State Routing

• Idea: distribute a network map• Each node floods costs c(u,v) to its neighbors• Given all costs, each node performs shortest

path (SPF) computation, e.g., using Dijkstra’s algorithm, between itself and all other nodes

20

Link-State vs. Distance-Vector

• Convergence– DV has count-to-infinity– DV often converges slowly (minutes) – Odd timing dependencies in DV

• Robustness– Route calculations a bit more robust under link-state. – DV algorithms can advertise incorrect least-cost paths

• Bandwidth Consumption for Messages• Computation• Security

21

OSPF: Salient Features

• Dijkstra, plus some additional features• Equal-cost multipath• Support for hierarchy: Inter-Area Routing

22

Example: Open Shortest Paths First (OSPF)

• Key Feature: hierarchy• Network’s routers divided into areas• Backbone area is area 0• Area 0 routers perform SPF computation

– All inter-area traffic travels through Area 0 routers (“border routers”)

Area 0

23

Example: IS-IS• Originally: ISO Connectionless Network Protocol (CLNP) .

– CLNP: ISO equivalent to IP for datagram delivery services

– ISO 10589 or RFC 1142

• Later: Integrated or Dual IS-IS (RFC 1195)– IS-IS adapted for IP

– Doesn’t use IP to carry routing messages

• OSPF more widely used in enterprise, IS-IS in large service providers

24

Area 49.001 Area 49.0002

Level-1Routing Level-2

Routing

Level-1Routing

Backbone

Hierarchical Routing in IS-IS

• Like OSPF, 2-level routing hierarchy – Within an area: level-1– Between areas: level-2– Level 1-2 Routers: Level-2 routers may also participate in L1 routing

25

Level-1 vs. Level-2 RoutingLevel 1 routing

– Routing within an area – Level 1 routers track links, routers, and end systems within L1 area– L1 routers do not know the identity of destinations outside their area. – An L1 router forwards all traffic for destinations outside its area to the

nearest L2 router within its area.

Level 2 routing– Routing between areas – Level 2 routers know the level 2 topology and know which addresses

are reachable via each level 2 router. – Level 2 routers track the location of each level 1 area. – Level 2 routers are not concerned with the topology within any level 1

area (for example, the details internal to each level 1 area). – Level 2 routers can identify when a level 2 router is also a level 1 router

within the same area. – Only a level 2 router can exchange packets with external routers located

outside its routing domain.

26

CLNS Addressing: “NSAPs”

• NSAP: Network-Service Attachment Point (a network-layer address)• All routers in the same area must have a common Area ID• System ID constraints

– Each node in an area must have a unique System ID – All level 2 routers in a domain must have unique System IDs – All NSAPs on the same router must have the same system ID. – All systems belonging to a given domain must have System IDs of the

same length in their NSAP addresses

System ID NSELAFI Variable length Area address

6 bytes 1 byte1 byte 1 - 12 bytes

Area ID Sys ID NSEL

27

ISIS on the Wire…

28

IS-IS Configuration on Abilene (atlang)

lo0 { unit 0 {

….family iso {

address 49.0000.0000.0000.0014.00; } …. }

isis { level 2 wide-metrics-only; /* OC192 to WASHng */ interface so-0/0/0.0 { level 2 metric 846; level 1 disable; }}

Only Level 2 IS-IS in Abilene

ISO Address Configured on Loopback Interface

29

IS-IS vs. OSPF

• Cisco ships OSPF in 1991• Cisco ships dual IS-IS in 1992• Circa 1995: ISPs need to run IGPs, IS-IS is

recommended due to the recent rewrite• IS-IS became very popular in late 1990s

– Deployed in most large ISPs (also Abilene)– Some ISPs (e.g., AOL backbone) even switched

30

Monitoring OSPF

• Challenge: How to get the OSPF Link State Advertisements (LSAs)?

31

Challenge #1: Capturing LSAs

• Wire-tap mode– Invasive– Dependent on Layer-2

• Host mode– Distribute LSAs over multicast– LSAR joins multicast group

• Full adjacency mode– Form high-cost adjacency with network

• Partial adjacency mode

32

Challenge #2: Dealing with Areas

• Problem: OSPF LSAs not advertised across area boundaries.

33

Today’s Papers: Alternative Intradomain Routing Mechanisms

• A Key Question: How to set weights in a shortest-path routing protocol?

• Problem: Link cost becomes a protocol knob, not a reflection of the topology

• Options:– Link-weight tuning– Set up circuits (MPLS, and route on different circuits)– Random perturbations on link weights– …